Authors: (including presenting author): :
Lam RPK(1)(2), Cheung KMH(2), Leung MKM(2), Chin JW(3), Wong KL(3), Qiu CC(4), Woo JCY(1), So RHY(4), Tsang TC(2), Rainer TH(1)(2)
Affiliation: :
(1) Department of Emergency Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong (2) Accident & Emergency Department, Queen Mary Hospital (3) PanopticAI (4) Department of Industrial Engineering and Analytics, Hong Kong University of Science and Technology
Keyword 1: :
Remote photoplethysmography
Keyword 4: :
Emergency department
Keyword 5: :
Artificial intelligence
Introduction: :
Remote photoplethysmography (rPPG) enables estimation of vital signs by using artificial intelligence (AI) algorithms to analyse the subtle changes of light reflected from the skin in facial videos captured by smartphone cameras.
Objectives: :
We aimed to evaluate: (1) the accuracy of a proprietary AI-based rPPG algorithm in contactless estimation of the heart rate (HR), respiratory rate (RR), SpO2, blood pressure (BP), and body temperature based on facial videos captured by smartphones; and (2) patient comfort and satisfaction with different measurement methods.
Methodology: :
A prospective observational cross-sectional study was conducted in Queen Mary Hospital Accident & Emergency Department. A convenience sample of adult patients of triage category 4 and 5 were recruited after informed written consent. Vital signs were measured manually by a research nurse using standard hospital equipment as reference standards. Facial videos of 25 seconds in length were recorded simultaneously using an iPhone 14 and analysed with a proprietary convolutional neural network algorithm for contactless vital sign estimation. The accuracy was evaluated using Pearson correlation coefficient (r) and root mean square error (RMSE). Patient satisfaction and comfort (100 mm visual analogue scale) were compared between measurement methods using the Wilcoxon signed-rank test.
Result & Outcome: :
From October to November 2024, 360 videos from 126 patients (79 women and 47 men; mean age 54 years) with sufficient signal quality were analyzed. Contactless and manual HR estimations had a high level of agreement (r 0.992, p< 0.05; RMSE 1.82 bpm). The respective values for other vital signs were: RR (r 0.589, p< 0.05; RMSE 3.48 breaths/min), SpO2 (r 0.173, p< 0.05; RMSE 1.65%), systolic BP (r 0.710, p< 0.05; RMSE 15.77 mmHg), diastolic BP (r 0.677, p< 0.05; RMSE 7.85 mmHg), and temperature (r 0.555, p< 0.05; RMSE 0.28°C). Patient comfort (p< 0.001) and satisfaction ratings (p< 0.001) were significantly higher with contactless measurements compared to manual measurements. This proof-of-concept project demonstrates the potential of AI-based rPPG in automating vital sign acquisition in clinical settings, which may save manpower and reduce infection risks in future pandemics.
